Prevalence and risk factors for CTX-M gram-negative bacteria ...

ORIGINAL ARTICLE

Prevalence and risk factors for CTX-M gram-negativebacteria in hospitalized patients at a tertiary care hospital in Kilimanjaro,Tanzania

Tolbert Sonda1,2 & Happiness Kumburu1,2& Marco van Zwetselaar1 & Michael Alifrangis3,4 & Blandina T. Mmbaga1,2 &

Ole Lund5& Frank M. Aarestrup6

& Gibson Kibiki2,7

Received: 23 November 2017 /Accepted: 18 January 2018# The Author(s) 2018. This article is an open access publication

AbstractEmergence and spread of extended spectrum beta-lactamase (ESBL)-producing gram-negative bacteria, mainly due toCTX-M, is a major global public health problem. Patients infected with ESBL-producing gram-negative bacteria have anincreased risk of treatment failure and death. We investigated the prevalence and risk factors for CTX-M gram-negativebacteria isolated from clinical specimens of patients hospitalized at a tertiary care hospital in Kilimanjaro, Tanzania.Isolated gram-negative bacteria from inpatients admitted at Kilimanjaro Christian Medical Centre (KCMC) betweenAugust 2013 and August 2015 were fully genome sequenced. The prevalence of ESBL-producing gram-negative bac-teria was determined based on the presence of blaCTX-M. The odds ratio (OR) and risk factors for ESBL-producing gram-negative bacteria due to CTX-M were assessed using logistic regression models. The overall CTX-M prevalence (95%CI) was 13.6% (10.1–18.1). Adjusted for other factors, the OR of CTX-M gram-negative bacteria for patients previouslyhospitalized was 0.26 (0.08–0.88), p = 0.031; the OR for patients currently on antibiotics was 4.02 (1.29–12.58), p =0.017; the OR for patients currently on ceftriaxone was 0.14 (0.04–0.46), p = 0.001; and the OR for patients with woundinfections was 0.24 (0.09–0.61), p = 0.003. The prevalence of ESBL-producing gram-negative bacteria due to CTX-M inthis setting is relatively low compared to other previous reports in similar settings. However, to properly stop furtherspread in the hospital, we recommend setting up a hospital surveillance system that takes full advantage of the availablenext-generation sequencing facility to routinely screen for all types of bacterial resistance genes.

Keywords ESBL . Gram-negative bacteria . Prevalence . CTX-M .Whole genome sequencing . Tanzania

* Tolbert [email protected]

Happiness [email protected]

Marco van [email protected]

Michael [email protected]

Blandina T. [email protected]

Ole [email protected]

Frank M. [email protected]

Gibson [email protected]

1 Kilimanjaro Christian Medical Centre, Kilimanjaro ClinicalResearch Institute, Moshi, Tanzania

2 Kilimanjaro Christian Medical University College, Moshi, Tanzania3 Centre for Medical Parasitology, Department of Immunology and

Microbiology, University of Copenhagen, Copenhagen, Denmark4 Department of Infectious Diseases, Copenhagen University Hospital,

Copenhagen, Denmark5 Centre for Biological Sequence Analysis, Technical University of

Denmark, Lyngby, Denmark6 DTU-Food, Centre for Genomic Epidemiology, Technical University

of Denmark, Lyngby, Denmark7 East African Health Research Commission, Bujumbura, Burundi

European Journal of Clinical Microbiology & Infectious Diseaseshttps://doi.org/10.1007/s10096-018-3196-8

Introduction

Existence of antibiotic resistance due to extended spectrumbeta-lactamase (ESBL)-producing gram-negative bacteria isa major global public health problem [1] and has been report-ed in all regions of the world [2–7]. Compared to patientsinfected with beta-lactam susceptible bacteria, patients infect-ed with ESBL-producing gram-negative bacteria put a greaterburden on health-care resources and have an increased risk oftreatment failure and poor outcomes including death [8, 9].High-income countries (HICs) have surveillance systems inplace to estimate the burden of bacterial infections due toESBL-producing gram-negative bacteria, and to determinerisk factors for acquisition of ESBL-producing gram-negativebacteria as well as the clinical outcomes associated with infec-tion [1, 10–12]. In Sub-Sahara Africa (SSA), data on ESBL-producing gram-negative bacteria epidemiology and risk fac-tors associated with ESBL-producing gram-negative bacteriainfection are scarce. When available, the data mostly origi-nates from hospital-based studies [13–18]. Several risk factorshave been documented to be associated with ESBL-producinggram-negative bacteria acquisition, including previous hospi-talization, previous use of antibiotics such as third-generationcephalosporins, hospital overcrowding, bed sharing whenhospitalized, and international travel [19–25].

In 2002, Blomberg et al. reported high proportions of in-fections due to ESBL-producing gram-negative bacteria at theNational Hospital in Dar es Salaam, Tanzania [26]. Since then,other reports have shown an increasing trend of ESBL-producing gram-negative bacteria infections in the same hos-pital [14, 27–29]. Similarly, studies conducted at a tertiaryhospital in Mwanza, Tanzania have documented an increasein ESBL-producing gram-negative bacteria infections [28, 30,31]. Like the centers in Dar es Salaam and Mwanza,Kilimanjaro Christian Medical Centre (KCMC) is a majorzonal tertiary healthcare facility, serving nearly one quarterof Tanzania’s estimated 53.5 million inhabitants [32]. Its set-ting however sets KCMC apart: the Kilimanjaro region inTanzania has among the highest international movements ofpeople in the country for tourism and business. The region hasfour distinct ecosystems inhabited by diverse communities ofminers, pastoralists, agrarians, and fishermen. Given the pub-lic health importance of ESBL-producing gram-negative bac-teria, the role that KCMC plays in guiding empirical treat-ment, and the need for therapeutic decisions to be based onguidelines which are derived from local epidemiological data[33], a study on ESBL-producing gram-negative bacteria inKCMC was found to be imperative. In our setting, little atten-tion is paid to gram-negative bacteria, although they havebeen reported in east Africa and elsewhere to be causes ofoutbreaks and to be main contributors of ESBL resistancedue to production of CTX-M [13, 34]. The aim of this studywas to determine the proportion estimates and risk factors for

ESBL-producing gram-negative bacteria due to CTX-Musingwhole genome sequences of all gram-negative bacteria iso-lated from clinical specimens at KCMC hospital.

Materials and methods

Study design, participants, and specimen collection

A hospital-based cross-sectional study was conducted at KCMCbetween August 2013 and August 2015. Geographically,KCMC is located in north-eastern Tanzania and serves as a zonalreferral hospital. The hospital has a bed capacity of 650 withapproximately 500 outpatients seeking medical services daily.

The study involved patients admitted in medical, surgical,and pediatrics wards who were suspected to have bacterialinfection. A written consent was obtained from each partici-pant or from parents or guardians of children before enrolmentinto the study. Clinical specimens collected for bacterial cul-ture were sputum, wound or pus swab, stool, and blood.

Bacterial culture and identification

Bacteria culture, isolation, and identification were performedaccording to in-house standard operating procedures as wellas the Clinical and Laboratory Standards Institute (CLSI)guidelines as described by Kumburu et al. [35].

Genomic DNA isolation, whole genome sequencing,and analysis

Bacterial genomic DNA (gDNA) was purified and its concen-tration determined using the Easy-DNA extraction kitInvitrogen® and the Qubit dsDNA Assay Kit Invitrogen®respectively. The gDNA library preparation was performedfollowing Nextera® XT DNA Sample Preparation Guide[36]. In brief, each gDNA was tagmented (tagged andfragmented) by the Nextera® XT transposome. Thetransposome simultaneously fragments the input DNA andadds adapter sequences to the ends. Thereafter followed alimited-cycle PCR amplification whereby indexes requiredfor cluster formation were added to each DNA piece. Theneach gDNA library was normalized to ensure equal represen-tation during sequencing. Equal volumes of the normalizedlibrary were combined, diluted in hybridization buffer, andheat-denatured prior to sequencing on the Illumina MiSeqplatform (Illumina, Inc., San Diego, CA). Whole genome se-quence reads were analyzed using bioinformatics tools avail-able on https://cge.cbs.dtu.dk/services/. For this report’spurpose, the analyses included de novo reads assembly,species identification, and determination of blaCTX-M

encoding ESBL.

Eur J Clin Microbiol Infect Dis

Collection of metadata and statistical analyses

Case report and laboratory result forms designed for this studywere used to collect metadata. Double data entry was per-formed in OpenClinica (OpenClinica LLC, MA USA). Stataversion 13 (StataCorp LP, Texas 77845 USA) was used for allstatistical analyses. The prevalence of resistance due to ESBL-producing gram-negative bacteria was calculated as the num-ber of gram-negative bacteria positive for blaCTX-M divided bytotal number of gram-negative bacteria sequenced. The prev-alence across levels of a categorical variable, such as age,gender, wards admitted, type of bed, hospitalization, antibi-otics use, disease conditions, and comorbidities, was com-pared using Chi-square test or Fisher’s exact test. Logisticregression was performed to determine risks of isolatingCTX-M gram-negative bacteria strains from patients. In themultivariate analysis, we included any variable whose p value≤ 10% and those that were considered key risk factors regard-less of their p values being > 10% in univariate analysis.Statistical significance of associations was decided based ona two-tailed p value and respective 95% confidence intervals.

Results

Study population

In total, 575 patients were included in this study (Table 1).These baseline data have also been presented before byKumburu et al. [35]. The median age in years (IQR) was 43(30–57), and 348 (60.6%) of the patients were males. Threehundred thirty nine (59.2%) had primary education level and271 (47.3%) were farmers. A total of 301 (52.3%) patientswere admitted into medical wards. One hundred six (18.4%)of all patients were on stretcher type beds. A total of 287(49.9%) specimens were wound and/or pus swabs.

A total of 263 (45.9%) patients had infected wounds, 81(14.1%) had pneumonia, and 20 (3.5%) had diarrhea. A totalof 60 (10.5%) were TB, 81(14.1%) were HIV, and 122(21.3%) were diabetic cases. Out of all participants, 412(71.9%) were on antibiotics at the time specimens were col-lected (Table 1).

Prevalence of CTX-M in gram-negative bacteria

A total of 590 specimens were collected in the followingdistribution: 56 were stool, 122 were sputum, 126 were blood,and 286 were wound or pus swabs. Of all the specimens, 249were culture-positive yielding 377 bacterial isolates that werefully sequenced. Of these, 287 bacterial isolates were gram-negative bacteria. The most prevalent gram-negative bacteriawere Proteus spp. (n = 48, 16.7%), Escherichia coli (n = 44,15.3%), Pseudomonas spp. (n = 40, 13.9%), and Klebsiella

spp. (n = 38, 13.2%). The CTX-M gene was found in 39 ofthe 287 gram-negative bacteria, yielding an overall prevalence(95% CI) of 13.6% (10.1–18.1) (Table 2). The prevalence inisolates from those admitted in surgical, medical, and ICUwards, respectively, were 5.0% (2.5–9.8), 9.4% (4.7–17.8),and 3.1% (0.4–19.8). Table 2 shows also crude odds ratios(OR) for different patient characteristics. The crude OR(95%CI) of CTX-M gram-negative bacteria in patients admit-ted in medical wards was significantly higher compared tothose from surgical wards; OR 3.62 (1.74–7.52), p = 0.001.Bacterial isolates from those sleeping in ward corridors/

Table 1 Demographic and clinical characteristics of the studypopulation

Patient characteristic Total

Number of patients 575 (100)

Median age in years (IQR) 43 (30–57)

Gender

Female 226 (39.4)

Male 348 (60.6)

Occupation

Farming 271 (47.3)

Employed 177 (30.9)

Other 125 (21.8)

Education level

No 130 (22.7)

Primary 339 (59.2)

Secondary/Above 104 (18.2)

Ward of admission

Surgical 257 (44.7)

Medical 301 (52.3)

Other 17 (3.0)

Location of patient

In Corridor/Stretcher 106 (18.4)

Inside room/Bed 469 (81.6)

Specimen collected

Blood 117 (20.3)

Sputum 120 (20.9)

Stool 51 (8.9)

Swabs (wound or pus) 287 (49.9)

Comorbid conditions

Wound 263 (45.9)

Pneumonia 81 (14.1)

Tuberculosis 60 (10.5)

Sepsis, septicemia 113 (19.7)

Diarrhea 20 (3.5)

Diabetes 122 (21.3)

HIV 81 (14.1)

Cancer 52 (9.1)

Currently on antibiotics 412 (71.9)

Eur J Clin Microbiol Infect Dis

stretchers had a prevalence of 7.4% (3.9–13.7), versus 4.8%(2.4–9.4) for those sleeping inside rooms/beds (Table 2).Prevalence of ESBL-producing gram-negative bacteria dueto CTX-M per different disease conditions was investigated(Table 3). Among patients with wound infections, the preva-lence was 4.2% (2.0–8.6) while among diarrhea cases, TBcases, and patients with sepsis, the prevalence was 2.0%(4.6–56.3), 6.7% (0.9–37.1), and 2.9% (0.7–11.3), respective-ly. Furthermore, ESBL-producing gram-negative bacteriaprevalence due to CTX-M in HIV and in diabetic cases was11.1% (3.5–30.0) and 8.5% (3.8–17.7), respectively.

Table 3 further documents crude ORs for different morbid-ities. Crude OR of CTX-M gram-negative bacteria in patientswith wounds was significantly lower compared to those whodid not have wounds; OR 0.28 (0.14–0.58), p = 0.001, whilethe crude OR of CTX-M gram-negative bacteria in patientswith diarrhea was significantly higher compared to those with-out; OR 11.0 (2.94–40.98), p = 0.001. The trend for OR of

infection with CTX-M gram-negative bacteria among HIVcases was higher compared to non-HIV cases; OR 2.46(0.96–6.30), p = 0.06.

We also investigated the prevalence of CTX-M gram-neg-ative bacteria in patients on different antibiotics (Table 4). Theprevalence of CTX-M gram-negative bacteria in patients on atleast one antibiotic was 6.8% (4.1–10.9). The OR of CTX-Mgram-negative bacteria when patients were on otherCeftriaxone as compared to those who were not showed thatthose on Ceftriaxone had a reduced risk; OR 0.34 (0.16–0.71),p = 0.004, while there was a trend that those on Ciprofloxacinhad an increased risk of CTX-M gram-negative bacteria; OR2.62 (0.88–7.83), p = 0.084.

Risk factors for CTX-M gram-negative bacteria

Table 5 shows adjusted OR (AOR) for various patient charac-teristics. After adjusting for other patient characteristics,

Table 2 Socio-demographiccharacteristics of patients fromwhich CTX-M gram-negativebacteria were isolated

Patient characteristics CTX-M gram-negative bacteria

Prevalence (95% CI) Crude OR (95% CI) P value

All CTX-M 13.6 (10.1–18.1) – –

Age (years)

0–35(n = 87) 4.6 (1.7–11.7) 1 (Ref) –

36–53(n = 84) 6.0 (2.5–13.6) 1.27 (0.53–3.01) 0.595

54 and above (n = 100) 8.0 (4.0–15.3) 1.12 (0.48–2.63) 0.786

Departmenta

SW (n = 159) 5.0 (2.5–9.8) 1 (Ref) –

MW (n = 85) 9.4 (4.7–17.8) 3.62 (1.74–7.52) 0.001

ICU (n = 32) 3.1 (0.4–19.8) 0.69 (0.15–3.20) 0.636

Hospital stay (days)

0–4 (n = 82) 7.3 (3.3–15.5) 1 (Ref) –

5–12 (n = 91) 6.6 (3.0–14.0) 1.05 (0.47–2.36) 0.911

13 and above (n = 86) 3.5 (1.1–10.4) 0.47 (0.18–1.25) 0.129

Gender

Female (n = 111) 4.5 (1.9–10.4) 1 (Ref) –

Male (n = 166) 7.2 (4.1–12.3) 0.91 (0.45–1.82) 0.783

History of hospitalization

No (n = 188) 6.4 (3.6–10.9) 1 (Ref) –

Yes (n = 99) 5.1 (2.1–11.7) 0.53 (0.24–1.16) 0.111

History of medication

No (n = 109) 3.7 (1.4–9.4) 1 (Ref) –

Yes (n = 178) 7.3 (4.3–12.2) 1.26 (0.62–2.58) 0.521

Current medication

No (n = 81) 8.6 (4.1–17.2) 1 (Ref) –

Yes (n = 206) 4.9 (2.6–8.8) 1.0 (0.47–2.12) 0.998

Patient location

Inside room/Beds (n = 166) 4.8 (2.4–9.4) 1 (Ref) –

Corridor/Stretcher (n = 121) 7.4 (3.9–13.7) 0.94 (0.48–1.88) 0.877

a sw surgical ward, mw medical ward, ICU intensive care unit

Eur J Clin Microbiol Infect Dis

several variables appeared to be statistically associated withCTX-M gram-negative bacteria. These variables include prior

admission, currently on antibiotics, currently on ceftriaxone,and wound infections. Those with a prior admission history

Table 3 Morbidity types andprevalence of CTX-M gram-negative bacteria

Patient characteristics CTX-M gram-negative bacteria

Prevalence (95% CI) Crude OR (95% CI) P value

Infected wounds

No (n = 109) 9.2 (4.9–16.3) 1 (Ref) –

Yes (n = 167) 4.2 (2.0–8.6) 0.28 (0.14–0.58) 0.001

Pneumonia

No (n = 261) 6.1 (3.8–9.8) 1 (Ref) –

Yes (n = 15) 6.7 (0.9–37.1) 1.61 (0.43–6.01) 0.475

Tuberculosis

No (n = 261) 6.1 (3.8–9.8) 1 (Ref) –

Yes (n = 15) 6.7 (0.9–37.1) 0.96 (0.21–4.44) 0.96

Sepsis

No (n = 220) 6.8 (4.1–11.0) 1 (Ref) –

Yes (n = 67) 2.9 (0.7–11.3) 0.56 (0.22–1.39) 0.211

Diarrhea

No (n = 266) 5.6 (3.4–9.2) 1 (Ref) –

Yes (n = 10) 2.0 (4.6–56.3) 10.97 (2.94–40.98) 0.001

Diabetes

No (n = 205) 5.4 (2.9–9.5) 1 (Ref) –

Yes (n = 71) 8.5 (3.8–17.7) 1.04 (0.48–2.26) 0.928

HIV

No (n = 249) 5.6 (3.3–9.3) 1 (Ref) –

Yes (n = 27) 11.1 (3.5–30.0) 2.46 (0.96–6.30) 0.06

Table 4 Antibiotic use andprevalence of CTX-M in gram-negative bacteria

Patient characteristics CTX-M gram-negative bacteria

Prevalence (95% CI) Crude OR (95% CI) P value

Currently on antibiotics

No (n = 54) 3.7 (0.9–13.9) 1 (Ref) –

Yes (n = 222) 6.8 (4.1–10.9) 1.35 (0.53–3.41) 0.529

Ciprofloxacin

No (n = 258) 6.2 (3.8–9.9) 1 (Ref) –

Yes (n = 18) 5.6 (0.7–32.1) 2.62 (0.88–7.83) 0.084

Ceftriaxone

No (n = 135) 8.1 (4.5–14.2) 1 (Ref) –

Yes (n = 141) 4.3 (1.9–9.2) 0.34 (0.16–0.71) 0.004

Cotrimoxazole

No (n = 267) 6.0 (3.7–9.6) 1 (Ref) –

Yes (n = 9) 11.1 (1.3–53.4) 3.31 (0.79–13.86) 0.101

Cloxacillin

No (n = 216) 6.0 (3.5–10.1) 1 (Ref) –

Yes (n = 60) 6.6 (2.5–16.6) 0.64 (0.25–1.61) 0.342

Metronidazole

No (n = 256) 5.9 (3.2–10.7) 1 (Ref) –

Yes (n = 20) 6.5 (3.1–13.1) 1.63 (0.52–5.17) 0.405

Eur J Clin Microbiol Infect Dis

showed a relatively reduced risk of CTX-M gram-negativebacteria compared to those with no prior history of admission;AOR 0.26 (0.08–0.88), p = 0.031. An increased risk of CTX-M gram-negative bacteria was observed for those currently onat least one antibiotic compared to those who were not on anyantibiotics; AOR 4.02 (1.29–12.58), p = 0.017. A decreasedrisk of CTX-M gram-negative bacteria was observed amongthose who were currently on Ceftriaxone compared to thosewho were not on Ceftriaxone; AOR 0.14(0.04–0.46), p =0.001 and for those with wound infections was 0.24 (0.09–0.61), p = 0.003. There is also seen a lower likelihood in pa-tients with wound infection than those without woundinfections.

Discussion

This survey investigated the prevalence and possible risk fac-tors for CTX-M gram-negative bacteria in hospitalized pa-tients in Kilimanjaro, Tanzania. It is the first report to investi-gate ESBL-producing gram-negative bacteria due to CTX-Min clinical specimens using whole genome sequencing and notbased on phenotypic testing. To the best of our knowledge, itis the first risk factor analysis of ESBL-producing gram-neg-ative bacteria infections in Tanzania. The overall prevalenceof CTX-M gram-negative bacteria in this study was 13.6%.This hospital-study’s findings are comparable with other hos-pital studies that had previously reported low ESBL-producing gram-negative bacteria prevalence in Gabon,15.0% [37], and Central Africa Republic, 4 to 19% [38].However, the report’s prevalence contrasts with 42% preva-lence in a review on general ESBL-producing gram-negativebacteria in East Africa hospitals [34], a 29.7% prevalencepreviously reported in Kilimanjaro by Kajeguka et al. [39],and 32.6% prevalence reported in Guinea-Bissau [22].Further, our findings contrast with other reports on ESBL-

producing gram-negative bacteria in hospitals outside theEast African region in that they had shown higher prevalencethan in this report: Cameroon 55% [19], Ghana 49% [23],Egypt above 45% [40], and South Africa 83% [41]. Therecould be several reasons for the observed low prevalence(13.6%) of CTX-M gram-negative bacteria, including the factthat the majority of patients were sampled while they had beenon antibiotics for a number of days since admission. The prev-alence of CTX-M gram-negative bacteria was observed todecrease with hospitalization days. The use of ceftriaxone inthe majority of inpatients at KCMC is a plausible explanationfor the observed low prevalence of CTX-M, and the decreasein prevalence over time. Moreover, the prevalence of CTX-Mat admission may be relatively high due to selection of ESBL-producing gram-negative bacteria in the community due tooveruse of antibiotics other than ceftriaxone. A comparisonof CTX-M resistance between patients on ceftriaxone versuspatients on other cephalosporins like ceftazidime or cefepimewould have been of great interest; however, the available datawas insufficient for this analysis.

Prevalence of CTX-M gram-negative bacteria appeared toincrease with age. The numbers of patients with age more than65 years in this study were low, so the comparison with otherpublications could not be done. Other reports [42–44] haveshown adults over 65 to be almost three times at risk of infec-tion with ESBL-producing gram-negative bacteria. Due totheir diminishing immunity and repeated hospitalization, thisgroup is more prone to infections including resistant bacterialstrains. However, there appears to be no statistical evidence tosupport the observed proportion differences in the risks of theCTX-M gram-negative bacteria across age groups.

We investigated whether or not sleeping on stretchers andsleeping in corridors as proxy indicators of overcrowdingwere associated with the CTX-M gram-negative bacteria. Itwas found that the prevalence of CTX-M gram-negative bac-teria in patients sleeping in ward corridors (on stretchers) asthe proper beds were all occupied and no space inside therooms were relatively high compared to their counterparts.However, sleeping on stretchers and sleeping in corridorswere not significantly associated with increased risk of isolat-ing CTX-M gram-negative bacteria.

Factors that showed statistically significant associationswith CTX-M gram-negative bacteria include prior admis-sion, currently on antibiotics, use of Ceftriaxone, andwound infections. This report shows that a large proportion(71.9%) of patients was already on antibiotics therapywhen specimens were collected for microbiological analy-ses. Prevalence of isolating CTX-M gram-negative bacte-ria from patients who were on at least one type of antibi-otics was higher (6.8%) than in those who were not (3.7%).The likelihood of isolating CTX-M gram-negative bacteriafrom those who were currently on antibiotics was found tobe four times higher than from those who were not on

Table 5 Multivariate logistic regression model showing adjusted ORsfor isolating CTX-M gram-negative bacteria

Patient characteristics AORa (95% CI) P value

Prior admission 0.26 (0.08–0.88) 0.031

Prior medication 1.50 (0.62–3.62) 0.368

Currently admitted in surgical ward 0.70 (0.28–1.76) 0.444

Patient inside the room 0.76 (0.32–1.78) 0.522

Over 4 days of hospitalization 0.72 (0.31–1.71) 0.462

Currently on antibiotic 4.02 (1.29–12.58) 0.017

Currently on Ciprofloxacin 0.84 (0.21–3.35) 0.804

Currently on Ceftriaxone 0.14 (0.04–0.46) 0.001

HIV positive 1.15 (0.37–3.55) 0.804

Wound infection 0.24 (0.09–0.61) 0.003

aAdjusted odds ratio

Eur J Clin Microbiol Infect Dis

antibiotics therapy. Reports have shown that the risk ofdeveloping antibiotics resistance in societies where antibi-otics are excessively used is high [44, 45] due to antimi-crobial resistance selection pressure. In KCMC, being areferral hospital, the patients coming to seek medical ser-vices are those who might have used antibiotics but treat-ment failed due to carriage of resistant strains prior to theirreferrals. Furthermore, given the fact that patients are re-ferred to our hospital from different healthcare facilities,the likelihood of patients to have been on antibiotics ishigh. Self-referred patients rarely have a paper record ofprior medications. Empirical therapy, self-medication, andover-the-counter medications are common in low-incomecountries (LIC) [46–48]. This experience-based therapypracticed by physicians in LIC is one of the factors fuellingemergence of resistance, particularly when unnecessarilyswitching to higher generations of antibiotics.

Ceftriaxone has been a drug of choice by many physi-cians in our setting due to its broad-spectrum activity. Over51% of the analyzed gram-negative bacteria were isolatedfrom patients who were on Ceftriaxone. The prevalence ofCTX-M gram-negative bacteria was 4.3% from patientswho were on Ceftriaxone, compared to 8.1% for thosewho were not. The likelihood of isolating CTX-M gram-negative bacteria from those who were currently onCeftriaxone appeared to be seven times lower than thosewho were not treated with Ceftriaxone. As self-medicationwith Ceftriaxone is uncommon due to its high cost, thissuggests that treatment with injectable Ceftriaxone in thehospital was effective against CTX-M gram-negativebacteria. Nevertheless, its prescription should be guidedby microbiological results.

A substantial proportion (34.5%) of CTX-M gram-neg-ative bacteria was isolated from patients who had a historyof admission (hospitalization). The prevalence of isolatingCTX-M gram-negative bacteria from patients with no his-tory of admission (hospitalization) was 6.4%, higher than5.1% in those with hospitalization history. However, thelikelihood of isolating CTX-M gram-negative bacteriafrom patients with history of admission was approximatelyfour times lower than their counterparts, consistent with arecent report [49] that showed a 32% lower risk of ESBL-producing gram-negative bacteria among previously hos-pitalized patients. It could be that during past admissions,these patients were put on Ceftriaxone doses that had suc-cessfully treated CTX-M gram-negative bacteria. This re-port suggests that Ceftriaxone could be effective againstCTX-M gram-negative bacteria. Furthermore, the observedhigh risk of CTX-M gram-negative bacteria among thosewho had no prior admission could be explained by the factthat they were carrying resistant CTX-M gram-negativebacteria from the community. However, these findingsare different from reports [20, 45, 50, 51] which found that

previous admission and use of cephalosporins increase therisk of CTX-M gram-negative bacteria.

A benefit of this study is its contribution to a largerproject in which we have been able to build an infrastruc-ture and implement next-generation sequencing (NGS)techniques for clinical diagnostics. For the current report,we have been able to analyze clinical specimens by doingwhole genome sequencing to identify CTX-M genes thatcode for ESBL. We acknowledge that several limitationsneed to be addressed. Firstly, wound and pus swab spec-imens make up the majority of samples studied.Microbiology results do not always represent the causeof the infection. Secondly, in this report, ESBL-producing gram-negative bacteria resistance was definedbased on the presence of the CTX-M gene. Despite itsimportant role in resistance, this restricted definitionmeans we may have underestimated the ESBL-producinggram-negative bacteria prevalence. Underestimation ofESBL-producing gram-negative bacteria resistance couldalso be attributed to sampling patients who were alreadyon antibiotics treatment. Thirdly, though our report high-lights some key aspects regarding prevalence and riskfactors for ESBL-producing gram-negative bacteria resis-tance due to CTX-M gram-negative bacteria that are im-portant in programming for antimicrobial control, thestudy was insufficiently powered to confidently draw con-clusions about all suspected variables, especially thosethat appeared contradicting. Fourth, this study was unableto differentiate between community-acquired andhospital-acquired ESBL-producing gram-negative bacteriaas clinical sampling was done during, rather than prior tohospitalization. Finally, the widespread use of antibioticsin our settings makes it difficult to ascertain whether ornot exposure to antibiotics preceded the emergency ofESBL-producing gram-negative bacteria resistance dueto CTX-M.

Conclusions

In conclusion, although the prevalence of ESBL-producinggram-negative bacteria due to CTX-M in this setting wasfound to be low, it is worthwhile to devise approaches aimedat containing the situation before it gets out of control. Toproperly stop their spread in the hospital, apart from directcontainment methods, we recommend setting up a hospitalsurveillance system that takes full advantage of the availableNGS facility to routinely screen for other resistant bacterialgenes.

Acknowledgements We thank all patients who consented to participatein this study, and the management of Kilimanjaro Christian MedicalCentre. We appreciate the efforts of the Data Management team of

Eur J Clin Microbiol Infect Dis

Kilimanjaro Clinical Research Institute, in particular Salim Semvua,Lilian Mboya, and Krisanta Wilhelm who were the data entry clerks.

Availability of data and materials Data are available on request to theauthors.

Authors’ contributions TS conceived the initial idea. FA, OL, and GKrefined the idea. TS and HK performed laboratory analyses. TS and MZanalyzed data and prepared manuscript draft. All authors read, revised,and approved the final manuscript.

Funding This study was supported by DANIDA through DanidaFellowship Centre award number DFC No. 12-007DTU.

Compliance with ethical standards

Ethical approval and participant’s consent This study was granted eth-ical approval by the KCMCResearch Ethics Committee and the NationalInstitute for Medical Research with approval numbers 893 andNIMR/HQ/R.8a/Vol.IX/2080 respectively. Awritten consent was obtain-ed from each participant or from parents or guardians of children beforeenrolment into the study.

Competing interests The authors declare that they have no competinginterests.

Open Access This article is distributed under the terms of the CreativeCommons At t r ibut ion 4 .0 In te rna t ional License (h t tp : / /creativecommons.org/licenses/by/4.0/), which permits unrestricted use,distribution, and reproduction in any medium, provided you give appro-priate credit to the original author(s) and the source, provide a link to theCreative Commons license, and indicate if changes were made.

References

1. World Health Organization (WHO) (2014) Antimicrobial resis-tance: global report on surveillance http://www.who.int/drugresistance/documents/surveillancereport/en/. Accessed 5Apr 2016

2. Souverein D, Euser SM, Herpers BL, Diederen B, Houtman P, vanSeventer M et al (2016) Prevalence, risk factors and molecularepidemiology of highly resistant gram negative rods in hospitalizedpatients in the Dutch region Kennemerland. AntimicrobialResistance & Infection Control 5:1–10. https://doi.org/10.1186/s13756-016-0107-6

3. Mehrgan H, Rahbar M (2008) Prevalence of extended-spectrumbeta-lactamase-producing Escherichia coli in a tertiary care hospitalin Tehran, Iran. Int J Antimicrob Agents 31:147–151. https://doi.org/10.1016/j.ijantimicag.2007.09.008

4. Zhang J, Zheng B, Zhao L,Wei Z, Ji J, Li L et al (2014) Nationwidehigh prevalence of CTX-M and an increase of CTX-M-55 inEscherichia coli isolated from patients with community-onset infec-tions in Chinese county hospitals. BMC Infect Dis 14:659. https://doi.org/10.1186/s12879-014-0659-0

5. CastanheiraM, Farrell SE, Krause KM, Jones RN, Sader HS (2014)Contemporary diversity ofβ-lactamases among Enterobacteriaceaein the nine U.S. census regions and ceftazidime-avibactam activitytested against isolates producing the most prevalent β-lactamasegroups. Antimicrob Agents Chemother 58:833–838. https://doi.org/10.1128/AAC.01896-13

6. Karabay O, Altindis M, Koroglu M, Karatuna O, Aydemir ÖA,Erdem AF (2016) The carbapenem-resistant Enterobacteriaceaethreat is growing: NDM-1 epidemic at a training hospital inTurkey. Ann Clin Microbiol Antimicrob 15:6. https://doi.org/10.1186/s12941-016-0118-4

7. Sid Ahmed MA, Bansal D, Acharya A, Elmi AA, Hamid JM, SidAhmed AM et al (2016) Antimicrobial susceptibility and molecularepidemiology of extended-spectrum beta-lactamase-producingEnterobacteriaceae from intensive care units at Hamad MedicalCorporation. Qatar Antimicrobial resistance and infection control5:4. https://doi.org/10.1186/s13756-016-0103-x

8. Esteve-Palau E, Solande G, Sánchez F, Sorlí L, Montero M, GüerriR et al (2015) Clinical and economic impact of urinary tract infec-tions caused by ESBL-producing Escherichia coli requiring hospi-talization: a matched cohort study. The Journal of infection 71:667–674. https://doi.org/10.1016/j.jinf.2015.08.012

9. Maslikowska JA, Walker SAN, Elligsen M, Mittmann N, PalmayL, Daneman N et al (2016) Impact of infection with extended-spectrumβ-lactamase-producing Escherichia coli or Klebsiella spe-cies on outcome and hospitalization costs. The Journal of hospitalinfection 92:33–41. https://doi.org/10.1016/j.jhin.2015.10.001

10. Muellner P, Pleydell E, Pirie R, Baker MG, Campbell D, Carter PEet al (2013) Molecular-based surveillance of campylobacteriosis inNew Zealand—from source attribution to genomic epidemiology.Euro Surveillence 18:1–7

11. Bathoorn E, Friedrich AW, Zhou K, Arends JP, Borst DM,Grundmann H, et al (2013) Latent introduction to the Netherlandsof multiple antibiotic resistance including NDM-1 afterhospitalisation in Egypt, August 2013. Euro Surveill. https://doi.org/10.2807/1560-7917.ES2013.18.42.20610

12. Knetsch C, Lawley T, Hensgens M, Corver J, Wilcox M, Kuijper E(2013) Current application and future perspectives of moleculartyping methods to study Clostridium difficile infections. EuroSurveill 18(4). Available online: http://www.eurosurveillance.org/ViewArticle.aspx?ArticleId=20381

13. Mshana SE, Hain T, Domann E, Lyamuya EF, Chakraborty T,Imirzalioglu C (2013) Predominance of Klebsiella pneumoniaeST14 carrying CTX-M-15 causing neonatal sepsis in Tanzania.BMC Infect Dis 13:466. https://doi.org/10.1186/1471-2334-13-466

14. Moyo SJ, Aboud S, Kasubi M, Lyamuya EF, Maselle SY (2010)Antimicrobial resistance among producers and non-producers ofextended spectrum beta-lactamases in urinary isolates at a tertiaryHospital in Tanzania. BMC research notes 3:348. https://doi.org/10.1186/1756-0500-3-348

15. Blomberg B, Manji KP, UrassaWK, Tamim BS, Mwakagile DSM,Jureen R et al (2007) Antimicrobial resistance predicts death inTanzanian children with bloodstream infections: a prospective co-hort study. BMC Infect Dis 7:1–14. https://doi.org/10.1186/1471-2334-7-43

16. Mshana SE, Matee M, Rweyemamu M (2013) Antimicrobial resis-tance in human and animal pathogens in Zambia, DemocraticRepublic of Congo. Mozambique and Tanzania : an urgent needof a sustainable surveillance system Annals of ClinicalMicrobiology and Antimicrobials 12:1. https://doi.org/10.1186/1476-0711-12-28

17. Meremo A, Mshana SE, Kidenya BR, Kabangila R, Peck R,Kataraihya JB (2012) High prevalence of non-typhoid Salmonellabacteraemia among febrile HIVadult patients admitted at a tertiaryhospital. North-Western Tanzania International Archives ofMedicine 5:1. https://doi.org/10.1186/1755-7682-5-28

18. Blomberg B, Mwakagile DSM, Urassa WK, Maselle SY,Mashurano M, Digranes A et al (2004) Surveillance of antimicro-bial resistance at a tertiary hospital in Tanzania. BMCPublic Health4:45. https://doi.org/10.1186/1471-2458-4-45

19. Lonchel CM, Melin P (2013) Extended-spectrum β-lactamase-producing Enterobacteriaceae in Cameroonian hospitals. Eur J

Eur J Clin Microbiol Infect Dis

Clin Microbiol Infect Dis 32:79–87. https://doi.org/10.1007/s10096-012-1717-4

20. Muvunyi CM, Masaisa F, Bayingana C, Mutesa L, MusemakweriA, Muhirwa G et al (2011) Decreased susceptibility to commonlyused antimicrobial agents in bacterial pathogens isolated from uri-nary tract infections in Rwanda: need for new antimicrobial guide-lines. The American journal of tropical medicine and hygiene 84:923–928. https://doi.org/10.4269/ajtmh.2011.11-0057

21. Lonchel CM, Meex C, Gangoué-Piéboji J, Boreux R, AssoumouM-CO, Melin P et al (2012) Proportion of extended-spectrum ß-lactamase-producing Enterobacteriaceae in community setting inNgaoundere. Cameroon BMC infectious diseases 12:53. https://doi.org/10.1186/1471-2334-12-53

22. Isendahl J, Turlej-Rogacka A, Manjuba C, Rodrigues A, Giske CG,Nauclér P (2012) Fecal carriage of ESBL-producing E. coli and K.pneumoniae in children in Guinea-Bissau: a hospital-based cross-sectional study. PLoS One 7:e51981. https://doi.org/10.1371/journal.pone.0051981

23. Feglo P, Adu-Sarkodie Y, Ayisi L, Jain R, Spurbeck RR, SpringmanAC et al (2013) Emergence of a novel extended-spectrum-β-lactamase (ESBL)-producing, fluoroquinolone-resistant clone of ex-traintestinal pathogenic Escherichia coli in Kumasi, Ghana. J ClinMicrobiol 51:728–730. https://doi.org/10.1128/JCM.03006-12

24. Obeng-Nkrumah N, Twum-Danso K, Krogfelt KA, Newman MJ(2013) High levels of extended-spectrum beta-lactamases in a ma-jor teaching hospital in Ghana: the need for regular monitoring andevaluation of antibiotic resistance. The American journal of tropicalmedicine and hygiene 89:960–964. https://doi.org/10.4269/ajtmh.12-0642

25. VonWintersdorff CJH, Penders J, Stobberingh EE, Lashof AMLO,Hoebe CJPA, Savelkoul PHM et al (2014) High rates of antimicro-bial drug resistance gene acquisition after international travel, theNetherlands. Emerg Infect Dis 20:649–657

26. Blomberg B, Jureen R, Manji KP, Tamim BS, Mwakagile DSM,Urassa WK et al (2005) High rate of fatal cases of pediatric septi-cemia caused by gram-negative bacteria with extended-spectrumbeta-lactamases in Dar es Salaam, Tanzania. J Clin Microbiol 43:745–749. https://doi.org/10.1128/JCM.43.2.745-749.2005

27. Ndugulile F, Jureen R, Harthug S, Urassa W, Langeland N (2005)Extended spectrum beta-lactamases among gram-negative bacteriaof nosocomial origin from an intensive care unit of a tertiary healthfacility in Tanzania. BMC Infect Dis 5:86. https://doi.org/10.1186/1471-2334-5-86

28. Mshana SE, Kamugisha E, Mirambo M, Chakraborty T, LyamuyaEF (2009) Prevalence of multiresistant gram-negative organisms ina tertiary hospital in Mwanza. Tanzania. BMC research notes 2:49.https://doi.org/10.1186/1756-0500-2-49

29. Manyahi J, Matee MI, Majigo M, Moyo S, Mshana SE, LyamuyaEF (2014) Predominance ofmulti-drug resistant bacterial pathogenscausing surgical site infections in Muhimbili National Hospital.Tanzania BMC research notes 7:500. https://doi.org/10.1186/1756-0500-7-500

30. Mshana SE, Falgenhauer L, Mirambo MM,Mushi MF, Moremi N,Julius R et al (2016) Predictors of blaCTX-M-15 in varieties ofEscherichia coli genotypes from humans in community settings inMwanza, Tanzania. BMC Infect Dis 16:1–9. https://doi.org/10.1186/s12879-016-1527-x

31. Nelson E, Kayega J, Seni J, Mushi MF, Kidenya BR, Hokororo Aet al (2014) Evaluation of existence and transmission of extendedspectrum beta lactamase producing bacteria from post-deliverywomen to neonates at Bugando Medical Center. Mwanza-Tanzania BMC Research Notes 7:279. https://doi.org/10.1186/1756-0500-7-279

32. United Nations, Department of Economic and Social Affairs,Population Division (2015). World population prospects:The2015 Revision, Key Findings and Advance Tables. Working

Paper No. ESA/P/WP.241. doi: https://doi.org/10.1017/CBO9781107415324.004

33. García-Tello A, Gimbernat H, Redondo C, Arana DM, Cacho J,Angulo JC (2014) Extended-spectrum beta-lactamases in urinarytract infections caused by Enterobacteria: understanding and guide-lines for action. Actas urologicas espanolas 38:678–684. https://doi.org/10.1016/j.acuro.2014.05.004

34. Sonda T, Kumburu H, van Zwetselaar M, Alifrangis M, Lund O,Kibiki G et al (2016) Meta-analysis of proportion estimates ofextended-spectrum-beta-lactamase-producing Enterobacteriaceaein East Africa hospitals. Antimicrobial Resistance & InfectionControl 5:1–9. https://doi.org/10.1186/s13756-016-0117-4

35. Kumburu HH, Sonda T,MmbagaBT, AlifrangisM, LundO, KibikiG et al (2017) Patterns of infections, aetiological agents and anti-microbial resistance at a tertiary care hospital in northern Tanzania.Trop Med Int Health 22:454–464. https://doi.org/10.1111/tmi.12836

36. Illumina Inc. Nextera® XT Library Prep Reference Guide[Internet]. Sample Preparation Guide 2016;1–28. doi: http://support.illumina.com/downloads/nextera_xt_sample_preparation_guide_15031942.html

37. Alabi AS, Frielinghaus L, Kaba H, Kösters K, Huson MAM, KahlBC et al (2013) Retrospective analysis of antimicrobial resistanceand bacterial spectrum of infection in Gabon. Central Africa BMCinfectious diseases 13:455. https://doi.org/10.1186/1471-2334-13-455

38. Bercion R, Mossoro-kpinde D, Manirakiza A, Le Faou A (2009)Increasing prevalence of antimicrobial resistance amongEnterobacteriaceae uropathogens in Bangui, Central AfricanRepublic. Journal of infection in developing countries 3:187–190

39. Kajeguka DC, Nambunga PP, Kabissi F, Kamugisha B, Kassam N,Nyombi B et al (2015) Antimicrobial resistance patterns of pheno-type extended spectrum beta-lactamase producing bacterial isolatesin a referral hospital in northern Tanzania. Tanzania Journal ofHealth Research 17. https://doi.org/10.4314/thrb.v17i3.%c

40. Zaki MES (2007) Extended spectrum beta-lactamases amonggram-negative bacteria from an Egyptian pediatric hospital: atwo-year experience. Journal of infection in developing countries1:269–274

41. Buys H, Muloiwa R, Bamford C, Eley B (2016) Klebsiellapneumoniae bloodstream infections at a South African children’shospital 2006–2011, a cross-sectional study. BMC Infect Dis 16:570. https://doi.org/10.1186/s12879-016-1919-y

42. Rodríguez-Baño J, Picón E, Gijón P, Hernández JR, Ruíz M, PeñaC et al (2010) Community-onset bacteremia due to extended-spectrum beta-lactamase-producing Escherichia coli: risk factorsand prognosis. Clinical infectious diseases : an official publicationof the Infectious Diseases Society of America 50:40–48. https://doi.org/10.1086/649537

43. Tumbarello M, Spanu T, Sanguinetti M, Citton R, Montuori E,Leone F et al (2006) Bloodstream infections caused by extended-spectrum-beta-lactamase-producing Klebsiella pneumoniae: riskfactors, molecular epidemiology, and clinical outcome.Antimicrob Agents Chemother 50:498–504. https://doi.org/10.1128/AAC.50.2.498-504.2006

44. Ben-Ami R, Rodríguez-Baño J, Arslan H, Pitout JDD, Quentin C,Calbo ES et al (2009) A multinational survey of risk factors forinfectionwith extended-spectrum beta-lactamase-producing entero-bacteriaceae in nonhospitalized patients. Clinical infectious dis-eases : an official publication of the Infectious Diseases Society ofAmerica 49:682–690. https://doi.org/10.1086/604713

45. Cornejo-Juárez P, Pérez-Jiménez C, Silva-Sánchez J, Velázquez-Acosta C, González-Lara F, Reyna-Flores F et al (2012)Molecular analysis and risk factors for Escherichia coli producingextended-spectrum β-lactamase bloodstream infection in

Eur J Clin Microbiol Infect Dis

hematological malignancies. PLoS One 7:e35780. https://doi.org/10.1371/journal.pone.0035780

46. van den Boogaard J, Semvua HH, Boeree MJ, Aarnoutse RE,Kibiki GS (2009) Sale of fluoroquinolones in northern Tanzania:a potential threat for fluoroquinolone use in tuberculosis treatment.J Antimicrob Chemother 65:145–147. https://doi.org/10.1093/jac/dkp413

47. Gwimile JJ, Shekalaghe SA, Kapanda GN, Kisanga ER (2012)Antibiotic prescribing practice in management of cough and/or di-arrhoea in Moshi municipality. Northern Tanzania: cross-sectionaldescriptive study The Pan African medical journal 12:103

48. Thriemer K, Katuala Y, Batoko B, Alworonga J-P, Devlieger H,Van Geet C et al (2013) Antibiotic prescribing in DR Congo: aknowledge, attitude and practice survey among medical doctorsand students. PLoS One 8:e55495. https://doi.org/10.1371/journal.pone.0055495

49. Al Yousef SA, Younis S, Farrag E, Moussa HS, Bayoumi FS, AliAM (2016) Clinical and laboratory profile of urinary tract infectionsassociated with extended spectrum B-lactamase producingEscherichia coli and Klebsiella pneumoniae. Ann Clin Lab Sci46:393–400

50. VardiM, Kochavi T, Denekamp Y, Bitterman H (2012) Risk factorsfor urinary tract infection caused by Enterobacteriaceae withextended-spectrum beta-lactamase resistance in patients admittedto internal medicine departments. Isr Med Assoc J 14:115–118

51. Park YS, Bae IK, Kim J, Jeong SH, Hwang S, Seo Y-H et al (2014)Risk factors and molecular epidemiology of community-onset ex-tended-spectrum β-lactamase-producing Escherichia coli bacter-emia. Yonsei Med J 55:467–475. https://doi.org/10.3349/ymj.2014.55.2.467

Eur J Clin Microbiol Infect Dis